New test version available - bugs may be fixed

[Brian the Fist]

Thanks to everyone who sent us their broken directories for analysis. I have gone through them all and there are one or two that still puzzle me. For the rest, it appears most problems are caused by people exiting improperly - this includes issuing kill signals in UNIX or closing the DOS text client by clicking the X in the upper right of the DOS Window. Some may also be a result of monitors such as dfGUI reading the filelist.txt at the same time as the client is trying to erase it (and rewrite it). And I believe I have identified a possible logic error which could account for some of the 'missing .val' messages but not certain on this one.

Anyhow, the code has been made more robust against all these situations and I believe most, but not necessarily all, of the problems people have been having should go away now.

Thus I'd request those of you, especially those who've been experiencing lots of problems with the latest client, please try out this new one so I know which bugs still exist, if any. You can grab it here:

ftp://ftp.mshri.on.ca/pub/distribfold/download/foldtrajlite_test1.zip

and it is for Windows only so far. Just overwrite your foldtrajlite.exe with the one in the ZIP (after backing it up first of course). You do not have to delete filelist.txt when you copy this over unless you were experiencing a problem the last time it ran. It will not fix already broken situations.

I can provide a Linux build if there is a demand too.

The only bug this definitely does NOT fix is the expr4parsevalue -490 error message, this one has stumped us so far.

Thanks and please provide your feedback here as to whether the new version is any better.

[IronBits]

Thanks Howard :thumbs: :cheers:
Put it on a dual amd, barton xp2800, xp2600, w2k/xp, no services

[gOhAsE]

Will run from now on for 2 days on a XP2600+ - then I'll be one week on vacation. :)

[GHOST]

How am I supposed to properly shut down the client in linux?

[IronBits]

cd /towheretheclientis
rm *.lock
:)

[FoBoT]

or press "Q"

[IronBits]

Good point! :)
I assumed he was running it in quiet mode -qt and,
in the background like a service with & ;)

[MgKnight]

Done in corrupted directory but I guess the corrupted filelist.txt file means all the work is lost forever eh? :(

Oh well, it is up and running again. :thumbs:

[GHOST]

yes i am running quiet. was thinking i needed a command for the terminal. but now i open distribfold folder and remove lock.
thanks guys.

[willebenn]

If the next update/changeover is close (maybe next week???) I can easily wait for the Linux version. If it's not that close to changeover time, then a Linux (ICC version) would be nice to test for a while before the change.

[Mpemba Effect]

Originally posted by GHOST
yes i am running quiet. was thinking i needed a command for the terminal. but now i open distribfold folder and remove lock.
thanks guys.

Create a rc-script to automatically launch df at bootup and automatically delete the lock file during the shutdown sequence. How you do this depends on your distro, heres how I did it in gentoo http://www.free-dc.org/forum/showthread.php?s=&threadid=1819

[PinHead]

Well I was pretty happy this afternoon as the 2K dualie made it past gen 0. Installed the patched file on deamon #2, NT4 boxen, and it lost track of the filelist.txt in under 2K folds.

You didn't by chance put that .net on the same machine that compiles vs6? Just curious if we need a newer mdac version, since I am assuming the data files are being written in foxpro.

[erk]

I have had several instances of filelist.txt corruption when the client under FreeBSD has shut itself down. The error.log said the client could not contact www.distributedfolding.org and the result was the client tried another server (the name I forget) then quit and left the filelist.txt with the wrong names in it. There were no comms problems at my end at the time.

[jonnyw]

should i use this file to replace the current one, even though i have had no errors.

I'm a bit of a believer of "if it aint broke, don't fix it", unless told to do otherwise

...anyone?

[Grumpy]

Dont change it :scared:

[Brian the Fist]

Ok folks, this conversation is not helping here. If you want to help, please grab the new version linked above (if you are using Windows of course) and please post complete, clear bug reports if you have any problems, old or new, with it. Thank You.

[jonnyw]

Originally posted by Grumpy
Dont change it :scared:

cheers, I wont :)

[PinHead]

filelist.txt:

CurrentStruc 0 36 127 54 1 5 9.942 -539.490 1078.340 202.040 10826657.000 1.650 3.100 2339.397 ------------------HHHHHHHHHHH--------HHHHH---------------HHHHH-----------------------HHHH-------
4e5fed0919e67e2cb19a2af1908d6569

error.log:

========================[ Aug 3, 2003 1:36 AM ]========================
FATAL ERROR: [003.001] {foldtrajlite2.c, line 1247} filelist.txt has been corrupted, cannot continue - please delete it and try again

files remaining in directory afterwards:
fold_0_handle_0_handle_protein_54.log.bz2
fold_0_handle_35_handle_protein_54.log.bz2
handle_0_handle_protein_53_0000006_min.val.bz2
handle_0_handle_protein_54_0000005.val
handle_0_handle_protein_54_0000036.val
handle_protein_54.trj


progress.txt:
Building structure 41 generation 54
9 until next generation
-1 generations buffered
Best Energy so far: 9.914

This is on the test version of foldtrajlite.exe

[Brian the Fist]

Thanks Pinhead, thats exactly what we need!:) Will get on it.
Could you possibly post your OS, if you were running as a service or not, and the flags in foldit.bat as well? I am puzzled by this error but someone has reported in the 'normal' version as well.

[jonnyw]

Originally posted by Brian the Fist
Ok folks, this conversation is not helping here. If you want to help, please grab the new version linked above (if you are using Windows of course) and please post complete, clear bug reports if you have any problems, old or new, with it. Thank You.

sorry, didn't read the title properly so didn't realise that it was only a test version (and not just an update of the .exe file)

will give it a go now :)

[MgKnight]

Yet another problem with this version. The program halted last night. It would not restart using dfGUI but I was able to upload the work by restarting it from the foldit.bat file. Once all the work was uploaded the program terminated with the MS error that it had generated errors and had to be shut down. The directory is intact with the exception of the change to delete the "-qt" switch so I could see what was happening. This is a WIN2K SP3 machine.

[PinHead]

For the Filelist.txt error posted earlier:

OS - NT4 Server
Running Console Mode
Switches -qt -rf

[MgKnight]

More info on my problem. Thinking perhaps there was a problem with dfGUI because it would not open on my screen I removed it and replaced it with a new copy. There was no change. I then attempted to restart the client using the foldit.bat file. It popped up a window saying it was "Making the Gen 171 Trajectory Distribution". Immediately following was a error message that errors had been made by the program and it would have to close.... Suggestions?

[Gortok]

To MgKnight:

Is this system overclocked?
We, at OCW, have found that the work done between generations, minimizing... making trajectory... etc. is the toughest on the system. If this is when you are experiencing problems, it may help to reduce the amount of FSB by 2-5 MHz and see if this helps any.

G

[MgKnight]

No it isn't. It is running pure stock.

[Brian the Fist]

I believe the crash at the end of creating a trajectory distribution has been fixed for the next release. It is not a hardware problem.